I’ve discovered increasing the TFTP block size significantly increases the speeds on both sides (>10MBps from TestA, ~1MBps from Client 1). I’d be OK with just increasing the blocksize for TFTP in FOG on Client 1 instead of resolving the TFTP issue between Client 1 and FOG Server. Both would be nice.
Does anyone know how to modify the block size for TFTP on the clients?
Hello,
I’ve recently set up FOG and I’ve noticed it takes about 30 minutes to download fog/kernel/bzImage and the file after that (I don’t remember exactly what it is). I upgraded the BIOS, that doesn’t seem to have changed anything, since I read that might cause the problem.
Here’s my setup:
[CODE]
TestA ========== FOG Server — Client 1
TestB ====/ ±- Client 2
±- Client 3
±- Client 4[/CODE]
The === is our corporate network and — is our isolated network. Everything is gigabit networking.
The FOG server acts as a NAT so that the clients can get out the corporate network. When I download bzImage using TFTP from the FOG Server to Client 1, it will download at about 60KBps. When I download the same file from the same server to TestA, it goes about 1-2 MBps. I thought it was a problem with the isolated network. But then I tried to download a different file: I downloaded a file (not using TFTP) from TestB to Client 1. And it went 1-2 MBps, too. So it seems that TFTP is only slow on the isolated side.
First Test: TFTP bzImage from FOG Server to Client 1: 60KBps
Second Test: TFTP bzImage from FOG Server to TestA: 1-2MBps
Third Test: SMB (Windows shares) from TestB to Client 1: 1-2MBps
What could possibly cause TFTP to be slow only for the isolated network?
I’ve noticed that the installation says I shouldn’t install FOG on a production machine because it installs a number of packages. I have a machine that I want to [I]add[/I] FOG too. Here’s what I’ve got:
[LIST]
[]Machine name: dcc
[LIST]
[]Apache 2, MySQL, DHCP, NAT with ~50 machines behind the NAT spread across 3 network interfaces on the private side, one on the public side.
[/LIST]
[]Machine name: dcgw-s1
[LIST]
[]DHCP, NAT with 15 worker machines behind a NAT (one public, one private interface)
[]Separate subnet from dcc
[/LIST]
[]Machine name: dcgw-s2
[LIST]
[]Same as dcgw-s1
[]Same subnet as dcgw-s1
[/LIST]
[*]Soon to be: dcgw-s3, dcgw-s4, dcgw-s5 (separate or same subnet as dcgw-s1)
[/LIST]
DCC is not something that we want to rebuild from scratch, but it also the machine that should probably have FOG on it, as the central coordinator for this little distributed cluster (the last C is for coordinator). It also runs the application that we use for managing all the workers and distributed jobs to those workers. We’re going to tie different images to different jobs, so we almost definitely want FOG on DCC.
There are some things we want:
[LIST]
[]Use an existing user (we already have a user for running apps like FOG’s)
[]Use existing installations and configurations for for MySQL, Apache, DHCP
[/LIST]
Is this possible? Or should I just start tearing apart the installation script and cobbling it back together?
Hello,
I’ve just found FOG and it appears to do almost everything that I need to control imaging on 70 machines now, growing to 115 in just a few weeks. We have these 115 machines spread out across different subnets with groups of 15 behind their own DHCP/NAT. I’ve been pleasantly surprised at how adaptable FOG seems to be.
However, there is one sticking point: I would like to be able to programmatically add a task to reimage a machine when it reboots. Is there any API (or not-too-hard way) to add a task like that?